Research On 3D Point Clouds Hole Repair With Hole Boundary Points Repulsion

3D point cloud hole repair technology is an important point cloud processing technology.Point cloud holes may be caused by instrument error or environment situation during scanning.It will seriously affect the subsequent processing accuracy of point clouds,especially when performing 3D reconstruction.The holes often cause the reconstruction accuracy to drop,the lack of details,and even the reconstruction failure.Regarding the holes repairing in 3D point clouds,there are still many problems such as too much manual interaction to determine the position of the hole,insufficient automation,and a large amount of calculation for hole repair.In response to these problems,this paper studies a high-automation and high-accuracy point cloud hole identification technology,and on this basis,studies a hole repair technology with fast calculation speed,which can maintain local geometric characteristics and does not change the original point cloud distribution.Finally,the repaired model is reconstructed to verify the effectiveness of the algorithm.This paper mainly researches on three contents:(1)3D point cloud preprocessing method based on normal informationThe scanned 3D point cloud contains a lot of redundant data,which brings a heavy burden to the subsequent calculation of the point cloud,and the original point cloud data often contains a lot of noise,which affects the accuracy of the subsequent point cloud processing,so this paper proposed a point cloud preprocessing method that can maintain the original structure of the point cloud model.This method first reduces the original point cloud data based on the point cloud reduction algorithm of the adaptive local projection operator.The reduced model not only greatly reduces the amount of data on the basis of maintaining the original model structural characteristics,but also reduces the noise.Afterwards,the normal vector optimization algorithm of anisotropic nearest neighbor search is used to optimize the normal vector,improve the accuracy of the normal vector,and lay the foundation for subsequent data processing.(2)3D point cloud hole selection based on curvature.There are often many holes in the 3D point cloud data,and the existing hole selection technology often requires manual intervention,and the judgment efficiency is affected.Therefore,this paper proposed a hole discrimination method that can automatically and accurately determine the position of the hole.This method uses the characteristics that the adjacent points of the hole boundary points are often distributed on the same side of the boundary point,and connects the target point to its adjacent point.If the maximum angle formed is greater than the empirical value,the point is judged as the hole boundary point.The addition of curvature constraints in the process can effectively reduce the error of judging the points data of sharp edges as the hole boundary points.(3)3D point cloud hole repairing based on repulsive force from neighborhood pointsThe current three-dimensional point cloud hole repair methods often need to triangulate the point cloud data and then process,or need to fit the local surface to repair the hole,which will consume a lot of calculation time,so this paper proposed a direct effect on point cloud data which is highly efficient and can maintain the original structural features of the model.This method uses the geometric relationship between the hole boundary point and its adjacent point to simulate a repulsive force.The adjacent point pushes the hole boundary point into the hole to complete the hole repair and introduces a loss function to adjust the position of the repair point so that the restored point cloud can still maintain local structural features.